Summary of the invention
Aspects and advantages of the present invention are partly stated in the following description, or can be apparent from this description, or learn by putting into practice the present invention.
For overcoming the problem of prior art, the invention provides a kind of method and system utilizing multi-view point video to carry out target following, by the method for path planning, reduce region in time or get rid of impossible camera, then in the region reduced or in possible camera, target location is accurately drawn by signature analysis, overcome the deficiency of blind search in existing multi-view point video specific objective tracking technique, improve the matching precision of multiple views specific objective, reduce Lou coupling and error hiding, drastically increase speed.
It is as follows that the present invention solves the problems of the technologies described above adopted technical scheme:
According to an aspect of the present invention, a kind of method utilizing multi-view point video to carry out target following is provided, comprises the following steps:
Obtain the distributing position of each camera, and set up the routing table of camera in conjunction with map, this routing table identifies any one camera and corresponding relation from this any one camera between next step camera of arriving of may moving;
Determine the camera at the current place of target, and preserve clarification of objective information, if target is still in current camera, then only in current camera, carry out target following; If target disappears in current camera, then read this routing table, search next step camera that may arrive of target according to this routing table, and in the video flowing of next step camera that may arrive, search target according to preserved characteristic information; If camera position is clearing, and find fall short for a long time in this routing table, then further according to the camera in this routing table, find the camera in its subpath table, to find target further.
According to one embodiment of present invention, when setting up routing table, if camera position is enclosure space, then whether directly sensiblely can set up routing table according between camera, the camera arrived that next step may be moved is listed in routing table.
According to one embodiment of present invention, when setting up routing table, if camera position is clearing, then set up routing table according to the distance of camera distance, all cameras in the region of pre-sizing are listed in routing table.
According to one embodiment of present invention, the region of this pre-sizing is a circle of delimiting centered by current camera.
According to one embodiment of present invention, when setting up routing table, if camera position is clearing and enclosure space intersection, then to enclosure space part, current camera is listed in routing table at the camera that subsequent time may arrive; To open space segment, all cameras in the region of pre-sizing are listed in routing table.
According to one embodiment of present invention, in this routing table, next step camera arrived that may move also comprises current camera.
According to one embodiment of present invention, this clarification of objective information comprise following at least one item: color characteristic, textural characteristics, shape facility, edge feature.
According to one embodiment of present invention, if target disappears in current camera, then according to this routing table, improve the sensitivity of next step camera that may arrive of target.
According to one embodiment of present invention, if the target of retrieving, then alarm, indicates the camera position at target place and the video of the camera at display-object place simultaneously.
According to another aspect of the present invention, a kind of system utilizing multi-view point video to carry out target following is provided, comprises:
Memory module, it stores the routing table of camera, this routing table is according to the distributing position of each camera and set up in conjunction with map, and this routing table identifies any one camera and corresponding relation from this any one camera between next step camera of arriving of may moving;
Tracking module, it determines the camera at the current place of target, and preserves clarification of objective information, if target is still in current camera, then only in current camera, carries out target following; If target disappears in current camera, then read the routing table in this memory module, search next step camera that may arrive of target according to this routing table, and in the video flowing of next step camera that may arrive, search target according to preserved characteristic information; If camera position is clearing, and find fall short for a long time in this routing table, then further according to the camera in this routing table, find the camera in its subpath table, to find target further.
The system utilizing multi-view point video to carry out target following can adopt the one or more further feature in method characteristic recited above equally.
The present invention is by the method for path planning, reduce region in time or get rid of impossible camera, then in the region reduced or in possible camera, target location is accurately drawn by signature analysis, overcome the deficiency of blind search in existing multi-view point video specific objective tracking technique, improve the matching precision of multiple views specific objective, reduce Lou coupling and error hiding, drastically increase speed.
By reading specification, those of ordinary skill in the art will understand the characteristic sum aspect of these embodiments and other embodiment better.
Embodiment
The present invention does path planning according to the position at current goal place, gets rid of in real time and can not appear in which camera after it, then analyzes the remaining camera video that may occur, determines the accurate location of target.Path planning problem is a classical combinatorial optimization problem, can regard as to construct some loops on given figure, makes these loops meet some specific conditions and constraint and the wastage in bulk or weight in these loops of route is minimum.
The method that the present invention utilizes multi-view point video to carry out target following comprises the following steps:
Obtain the distributing position of each camera, and set up the routing table of camera in conjunction with map, this routing table identifies any one camera and corresponding relation from this any one camera between next step camera of arriving of may moving;
Determine the camera at the current place of target, and preserve clarification of objective information, if target is still in current camera, then only in current camera, carry out target following; If target disappears in current camera, then read this routing table, next step camera that may arrive of target is searched according to this routing table, and in the video flowing of next step camera that may arrive, search target according to preserved characteristic information, if camera position is clearing, and find fall short for a long time in this routing table, then further according to the camera in this routing table, find the camera in its subpath table, to find target further.
Wherein, this clarification of objective information can adopt the combination of color characteristic, textural characteristics, shape facility, edge feature or these features.If target disappears in current camera, then according to this routing table, improve the sensitivity of next step camera that may arrive of target.If the target of retrieving, then alarm, indicates the camera position at target place and the video of the camera at display-object place simultaneously.If the target of not retrieving, then according to this routing table, proceed to search in the video flowing of next step camera that may arrive.
In one embodiment, first obtain the map of camera and map is analyzed, determine that camera position is enclosure space or clearing, if clearing then carrys out path planning according to the distance between camera, if enclosure space then according between camera whether can directly sensible come path planning.Its concrete steps are as follows:
1) camera position map obtains;
2) camera map path is analyzed;
3) camera continuous surveillance, has judged whether that anomalous event occurs;
4) target of anomalous event is followed the tracks of, simultaneously alarm;
5) if target disappears, exit.
The acquisition of the first step, camera position map
Obtain the situational map of camera, obtain actual map simultaneously, situational map and actual map are combined, be convenient to follow-uply belong to clearing to camera or enclosure space is done to judge.
Second step, camera map path are analyzed
1) clearing is belonged to each camera or belonging to enclosure space indicates.
Label 1-14 in label a1-a20 and Fig. 2 in Fig. 1 is for being camera.The camera being a1, a2, a3, a4 as ID in Fig. 1 is in clearing, ID is the intersection that the camera of a5 is then in clearing and enclosure space, if have passed through ID is that the camera of a5 enters in building, as shown in Figure 2, then all cameras all belong to the camera of enclosure space in building.
2) if the building closed, then carry out path planning according to the layout in the position of each camera and building, hypothetical target, at current camera, is analyzed next step camera position that may reach and sets up routing table.
As in Fig. 1, the position at current pedestrian target place is clearing; As shown in Figure 2, the position at current pedestrian target place is enclosure space.
3) if open space, then the distance according to camera distance sets up routing table.
As shown in Figure 1, the camera identification (ID) at current pedestrian target place is a2, the present invention delimit a region (as shown in phantom in fig. 1) centered by camera a2, the inside includes totally 5 cameras of camera a1, a2, a3, a4, a5, these 5 cameras are exactly its subsequent time position that may arrive is camera position, as the routing table of camera a2; After target arrives camera a3, delimit a region (as shown in phantom in fig. 1) with camera a3, the inside includes camera a2, a3, a4, a6, a7, as the routing table of camera a3, by that analogy.
But this routing table is most important and the routing table searched at first, if find fall short for a long time in this routing table, consider in open environment and have many blind areas to cause camera to monitor, then need further according to the camera ID in current path table, find the camera in its subpath table, to find target further.As shown in Figure 1, current pedestrian target is in camera 2, and so according to its routing table, follow-up needs find target in a1, a2, a3, a4, a5 totally 5 cameras for ratio.When not finding target for a long time, then find in subpath table further, the routing table of such as camera a3 is a2, a3, a4, a6, a7, and it is exactly the subpath table of camera a2, by that analogy.
4) if enclosure space, setting up routing table according to the path in building.
As shown in Figure 2, the camera ID at current pedestrian target place is 7, next step camera that may arrive of camera 7 is 5,6,13, so 5,6,7,13 is exactly the routing table of camera 7, here camera 7 is also extremely important as the part inside path, because target leaves camera likely again get back to this camera later.
Have 14 cameras in fig. 2, these 14 cameras are arranged at respectively can the porch of traffic areas or corner.As can be seen from Table 1, if target is positioned at camera 4 place, the position that its subsequent time may arrive is camera 2,4,8,10 place; After target arrives camera 10, the position that its subsequent time may arrive is camera 4,9,10,11 place; By that analogy, after target way camera 13, the position that its subsequent time may arrive is camera 7,12,13,14 place.
The path planning table of specific objective in table 1 building
Video camera ID |
Next video camera ID |
1 |
(1)、2、3 |
2 |
1、(2)、4 |
3 |
1、(3)、5、6 |
4 |
2、(4)、8、10 |
5 |
3、(5)、6、7 |
6 |
3、5、(6)、7、14 |
7 |
5、6、(7)、13 |
8 |
4、(8) |
9 |
(9)、10、12 |
10 |
4、9、(10)、11 |
11 |
10、(11)、12 |
12 |
9、11、(12)、13 |
13 |
7、12、(13)、14 |
14 |
6、13、(14) |
3rd step, capture anomalous event
1) when determining that event occurs, the camera position ID at target place
By the characteristic matching of moving object detection and behavioural analysis or intended target (as important persons or associated vehicle), determine its camera ID when event occurs, preserve the correlated characteristic information of this event target, for supervise is prepared simultaneously.
2) sign of current camera whether in building is read
The positional information of current camera is divided into three kinds: clearing, enclosure space, clearing and enclosure space have a common boundary.
3) according to camera current location read path table
Clearing: according to the area size of in advance setting using cameras all in this region as routing table.
Enclosure space: read current camera may to arrive camera routing table at subsequent time according to the path in building.
Clearing and enclosure space have a common boundary: be then read current camera may to arrive camera routing table at subsequent time to enclosure space inside; Then read camera routing tables all in this region according to the region delimited to clearing, above-mentioned two tables be exactly comprehensively the routing table of this camera.
4th step, target following
1) if target is still in current camera, then only in current camera, target analysis and tracking is carried out.The object matching of moving object detection and tracking and feature based can be combined, realize accurate tracking.
2) if target disappears in current camera, then according to routing table, analysis is carried out to next step camera that may arrive and find target.
5th step, alarm
1) if there is anomalous event to occur, then alarm, indicates camera position and display video simultaneously.
As shown in Figure 3, the target following flow process based on path planning of the present embodiment is as follows:
The video image that the 101. pairs of some cameras continue is analyzed, and judges whether that event occurs;
If 102. judge whether that event occurs, then enter step 103;
The position of 103. calculating cameras, and retain the feature of current goal;
104. pairs of targets are followed the tracks of;
105. judge target whether still in current camera; If, then enter step 104; If not, then enter step 106;
106. judge current camera whether in enclosed environment; If yes then enter step 107-110; Otherwise, enter step 111-114;
107., according to current camera position read path table, judge the position that next step target may arrive;
108. improve the sensitivity alarm that may arrive camera
109. judge whether to find target, if so, then enter step 104, otherwise enter step 110;
110. pairs of cameras that may arrive proceed to analyze;
111. according to current camera position initial setting region, carries out analysis find target to cameras all in this region;
112. improve the sensitivity of camera in regions and alarm;
113. judge whether to find target, if so, then enter step 104, otherwise enter step 114;
114. pairs of cameras that may arrive proceed to analyze, or enlarged area search.
As shown in Figure 4, the present invention provides a kind of system utilizing multi-view point video to carry out target following simultaneously, comprise: memory module and tracking module, this memory module stores the routing table of camera, this routing table is according to the distributing position of each camera and set up in conjunction with map, and this routing table identifies any one camera and corresponding relation from this any one camera between next step camera of arriving of may moving; The camera at this current place of tracking module determination target, and preserve clarification of objective information, if target is still in current camera, then only in current camera, carry out target following; If target disappears in current camera, then read the routing table in this memory module, search next step camera that may arrive of target according to this routing table, and in the video flowing of next step camera that may arrive, search target according to preserved characteristic information; If camera position is clearing, and find fall short for a long time in this routing table, then further according to the camera in this routing table, find the camera in its subpath table, to find target further.
The system utilizing multi-view point video to carry out target following can adopt the further feature in specifically described embodiment above equally.
The present invention is according to the camera map of setting, path planning is made in real time in current location, target can be found in the border circular areas of setting at open space, in the space of the relative closures such as building, then determine that the next camera that may arrive is searched for according to path, the position that can not occur is got rid of, and then accurate judgement is made in remaining possible position, not only substantially increase speed, and decrease error hiding.
The present invention is by the method for path planning, reduce region in time or get rid of impossible camera, then in the region reduced or in possible camera, target location is accurately drawn by signature analysis, overcome the deficiency of blind search in existing multi-view point video specific objective tracking technique, improve the matching precision of multiple views specific objective, reduce Lou coupling and error hiding, drastically increase speed.
Above with reference to the accompanying drawings of the preferred embodiments of the present invention, those skilled in the art do not depart from the scope and spirit of the present invention, and multiple flexible program can be had to realize the present invention.For example, to illustrate as the part of an embodiment or the feature that describes can be used for another embodiment to obtain another embodiment.These are only the better feasible embodiment of the present invention, not thereby limit to interest field of the present invention that the equivalence change that all utilizations specification of the present invention and accompanying drawing content are done all is contained within interest field of the present invention.